Микроконтроллеры

Там в файлах справки кое-что есть поверхностно:  The VNC2 architecture supports 8, 16 and 32 bit operations on memory addresses without registers. Operations of all sizes involving only data stored in RAM are fast and efficient. Loading 8, 16 and 32 bit values into memory is achieved in a single operation. Most operations can take one immediate value of 8, 16 or 32 bit size. Using 16 and 32 bit data value will, in practice, result in larger code though. Signed arithmetic is fully supported by hardware. Although conversions between signed values of differing sizes is performed in software will always generate larger and slower code than unsigned Intermediate values in calculations are assumed to be of type unsigned integer (32 bits). There is no floating point support in hardware. Signed modulus operations are not supported. Any modulus operation will be performed as if both dividend and divisor are unsigned.